Wrong results for log standard error

[statkwon]

Thanks for providing wonderful benchmark for XAI methods.

But there's a slight issue in the provided codes. When I evaluate metrics using evaluate_metrics.py, I cannot get the accurate results for log standard error.

For example, the scale of log standard error for RIS below seems awkward.

Model: lr, Data: adult, Explainer: control (1/16)1s total, 1s on model, 1s on dataset)
PRA: 0.501 ± 0.003

RC: 0.003 ± 0.009

FA: 0.188 ± 0.006

RA: 0.073 ± 0.005

SA: 0.093 ± 0.004

SRA: 0.034 ± 0.003

PGU: 0.130 ± 0.003

PGI: 0.043 ± 0.002

RIS: 420764.219 ± 1391.567
log(RIS): 12.950 ± 7.238

I made PR #31 to fix this issue. Please check this.